The present invention relates to content-aware dynamic network resource allocation in a network such as the Internet and in one embodiment to content-aware dynamic optical bandwidth allocation.
The Internet is a data transport network including legacy networks and optical networks. The legacy networks exist at the outer edge of the Internet and use copper wires to connect client systems and electronic routers/switches. The optical networks form the backbone of the Internet and use fibers to connect optical cross connects or switches. Between the legacy networks and optical networks are edge devices, which are electro-optic.
Well-developed optical transport technology brings an ever-increasing amount of bandwidth to the Internet. However, with conventional network technology, end clients have little ability to exploit optical network resources for their own purposes. As a result, there is abundant bandwidth available in the backbone of the Internet. In fact, a number of emerging Internet applications such as storage area network and streaming media intend to take greater advantage of the abundant bandwidth. These applications are dynamically initiated and provide a variety of content data over the IP protocol.
More specifically, conventional network technology has the following drawbacks.
1. Static Provisioning of Optical Links
Conventional network resource provisioning establishes fixed links to connect customer networks. For example, the optical provisioning is static, fixed bandwidth, and usually takes a long time, e.g., on the order of a month, because it involves inter-network-provider service negotiations and is accomplished manually.
2. Cost of the Optical Link Bandwidth is not Divided Among Many Users and Provisioning Bandwidth is not Flexible
Conventionally, provisioning an optical link (e.g., OC-48) means that a customer owns the full link bandwidth (i.e., 2.5 Gbps). This customer is fully responsible for the cost of the optical link. The cost is significant and the optical link bandwidth use is often inefficient. On the other hand, a number of users can share an optical link but they are not guaranteed a specified portion of the bandwidth.
3. Existing Signaling Protocols not Supported by Optical Gear
Applications can send their bandwidth requirements to the network using existing Internet signaling protocols such as Resource Reservation Protocol (RSVP). These protocols are at the Internet protocol (IP) layer, i.e., the layer 3 (L3), of the ISO Open System Interconnection reference model and require hardware support at network devices. However, optical devices perform data transport at the physical (L1) or the link (L2) layer and thus application signals are not processed in optical networks.
4. Optical Bandwidth Provisioning is not Aware of the Content of Application Traffic
Conventional bandwidth provisioning is based on the TCP/IP characteristics of traffic flow, which include IP protocol types, source and destination IP addresses, and TCP/UDP source and destination port numbers of traffic packets. Such provisioning is limited for L4 or higher-layer content differentiation because a client may use applications that deal with multiple content traffic streams at the same time. For example, audio and video applications employ different types of content and have different bandwidth requirements. On the other hand, optical networks do not support content differentiation because optical devices do not process IP packets.
Prior attempts to solve the problems described above include the following.
1. Intserv/RSVP
RSVP is the Intserv ReSerVation Protocol under the Internet Engineering Task Force (IETF) and is used by applications to signal the network for bandwidth reservations for IP traffic. However, RSVP is thought to be not scalable because backbone routers cannot maintain the flow status for all reserved traffic passing by. In addition, optical gear does not accept RSVP signals from end applications because RSVP is an L3 IP protocol.
2. ATM
Within an asynchronous transfer mode (ATM) network, applications can invoke the ATM user-network interface (UNI) to establish virtual circuits with particular bandwidth assignments. However, the ATM UNI is not applicable for non-ATM applications.
3. GMPLS
GMPLS (Generalized MultiProtocol Label Switching) is a known protocol of traffic path establishment for next-generation optical network. GMPLS is applied with the emerging ASTN (Automatic Switch Transport Network) technology. However, GMPLS does not support granular bandwidth requests from individual clients nor does it allocate bandwidth based on the content of application traffic.
Thus, there remains a need to more fully and effectively exploit the abundant bandwidth existing at optical networks.